Novelty-based Multiobjectivization
Résumé
Novelty search is a recent and promising approach to evolve neuro-controllers, especially to drive robots. The main idea is to maximize the novelty of behaviors instead of the efficiency. However, abandoning the efficiency objective(s) may be too radical in many contexts. In this paper, a Pareto- based multi-objective evolutionary algorithm is employed to reconcile novelty search with objective-based optimization by following a multiobjectivization process. Several multiobjectivizations based on behavioral novelty and on be- havioral diversity are compared on a maze navigation task. Results show that the bi-objective variant “Novelty + Fitness” is better at fine-tuning behaviors than basic novelty search, while keeping a comparable number of iterations to converge.
Domaines
Robotique [cs.RO]
Origine : Fichiers produits par l'(les) auteur(s)
Loading...